git gc can require using a *huge* amount of space

From: "Theodore Ts'o" <tytso@mit.edu>
Date: 2016-06-15 22:45:16

One of the things which I noticed which can be a bit of a misfeature is
if you are trying to save disk space by running "git gc", it can
sometimes take *substantially* more disk space.  For example, I ran "git
reflog --all --expire=0 --expire-unreachable=0" on a repository that had
linux-next as one of its remotes.  Before I ran "git gc --prune" the
repository was packed, and its .git/objects was a bit over 50 megs
(there was an alternates file).  While I was running the "git gc", it
was busy ejecting all of the objects that were no longer reachable
separate files, and the size of .git/objects ballooned to over 200 megs,
before settling in at 150 megs or so after "git gc" completed.  After I
ran "git prune", the size of .git/objects shrank back down to 48 megs.

That kind of swing in disk usage seems a bit excessive....  maybe
instead of pushing out all of the no-longer-reachable objects into
separate files, they could be put in a separate pack file (which the
user could then delete much more quickly than it would take "git prune"
to run)?

     	     	      	   	       	     - Ted
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help